McKinsey’s latest State of AI numbers say what a lot of operators already feel in their gut: pilots are cheap to run and easy to celebrate, but scaling AI is where companies actually fall behind. 88% of organizations now use AI somewhere in the business. Nearly two-thirds still haven’t scaled it across the enterprise. That’s the gap costing companies real money.
The Pilot Trap Most Companies Fall Into
A pilot feels like progress. Someone picks a use case, gets a quick win, shows it off at a town hall. Then it stalls.
McKinsey found something that should worry a lot of leadership teams: high performers are 2.8 times more likely to redesign workflows around AI instead of bolting a tool onto the old process. That’s the actual split. Not who has access to the best model, but who’s willing to change how work gets done.
The same pattern shows up with AI agents. 62% of organizations are experimenting with them. Only 23% have scaled even one. The bottleneck usually isn’t the model itself. It’s everything wrapped around it.
Why Workflow Redesign Beats Tool Adoption
A sharper model doesn’t fix a broken handoff between sales and ops. It just automates the broken part faster. McKinsey’s research keeps circling back to one finding: companies with real EBIT impact are the ones rebuilding workflows, not the ones with the biggest AI budget line.
That’s a harder pitch internally. Redesigning a process means retraining people and admitting the org chart doesn’t match how work actually flows now. A pilot never asks anyone to do that.
Heavy spending alone doesn’t predict success, either. Which is the part finance teams don’t love hearing when they’re expecting a clean ROI curve.
What It Takes to Actually Scale
Companies that pull this off share a few habits. They tie each AI initiative to a specific business outcome instead of letting it run as a standalone experiment. They fix their data architecture before automating decisions on top of messy data. And someone owns the outcome, not just the rollout.
Agentic AI raises the stakes. One agent running on fragmented data breaks down fast. Multi-agent systems make it worse, since coordination errors stack on top of each other across the chain.
If your company has stuck with a structured CMS instead of chasing platform migrations for years, that consistency is actually a head start. Clean, well-organized data is the thing scaling depends on.
Conclusion: Fewer Pilots, More Follow-Through
McKinsey’s surveys keep landing on the same answer. The companies winning with AI aren’t running the most pilots. They’re the ones rebuilding how work happens around it.
